Item — HN Reader

It could have been better if the entire codebase could always be provided to AI as the context. Otherwise, specifying exactly what you want is one step away from just doing it yourself.

> As a manager, AI is really nice to get a summary of how everything is going at the company and what tasks everyone is working on and the status of the tasks, instead of having refinement meetings to get status updates on tasks.

I do not understand why they are not marketing some "GPT Middle Manager" to the executive boards so that they could cut that fat. Surely that is a huge untapped cost-cutting potential?

3 comments

Cannot be worse than almost all human managers, so agreed. I am a terrible manager myself; i'm a good ceo/cto, making profits and keeping things running for almost no money, but managing i'm terrible at. And I haven't seen many who couldn't be replaced by a piece of cardboard. There are exceptions, but AI's can just as well do terrible team management while keeping their upper managers/c-levels busy with nonsense documents as is the standard for humans too. Indeed yesterday I wrote on HN that this is what LLMs are VERY good for; generating ENORMOUS piles of paper to give to all types of (middle) management to make them feel valued.

The obvious next step is where you can easily put new knowledge inside the parameters of the model itself.

I want the AI to know my codebase the same way it knows the earth is round. Without any context fed to it each time.

Instead we have this weird Memento-esque setup where you have to give it context each time.

they are already doing that. full-steam on. just look at HSBC made their goal to layoff middle-managers. and so does Google, Microsoft, others.

We already had that cycle with agile. I predict half baked models, chaos, then backslash with hiring even more managers combining models and management into one large innefective bundle.

The ones profiting the most will be consultancies designed to protect the upper management reputation.